Menu
▾
▴
[Teem-commit] SF.net SVN: teem:[7236] teem/trunk/python/cffi/exult .py
|
From: <kin...@us...> - 2024-09-06 08:00:31
|
Revision: 7236
http://sourceforge.net/p/teem/code/7236
Author: kindlmann
Date: 2024-09-06 08:00:29 +0000 (Fri, 06 Sep 2024)
Log Message:
-----------
unfortunately a mix of (1) new formatting with a new version of blue, and (2) changing how #defines in header files are processed: from a list of strings defined, to: a dict mapping from name to value, and a modest effort to apply those substitutions prior to sending things to cffi.cdef()
Modified Paths:
--------------
teem/trunk/python/cffi/exult.py
Modified: teem/trunk/python/cffi/exult.py
===================================================================
--- teem/trunk/python/cffi/exult.py 2024-09-05 20:51:26 UTC (rev 7235)
+++ teem/trunk/python/cffi/exult.py 2024-09-06 08:00:29 UTC (rev 7236)
@@ -113,7 +113,17 @@
},
'ten': {
'expr': False,
- 'deps': ['echo', 'limn', 'gage', 'dye', 'unrrdu', 'ell', 'nrrd', 'biff', 'air'],
+ 'deps': [
+ 'echo',
+ 'limn',
+ 'gage',
+ 'dye',
+ 'unrrdu',
+ 'ell',
+ 'nrrd',
+ 'biff',
+ 'air',
+ ],
},
'elf': {
'expr': True,
@@ -133,7 +143,16 @@
},
'mite': {
'expr': False,
- 'deps': ['ten', 'hoover', 'limn', 'gage', 'ell', 'nrrd', 'biff', 'air'],
+ 'deps': [
+ 'ten',
+ 'hoover',
+ 'limn',
+ 'gage',
+ 'ell',
+ 'nrrd',
+ 'biff',
+ 'air',
+ ],
},
'meet': {
'expr': False,
@@ -193,8 +212,8 @@
except Exception as exc:
raise RuntimeError(f'{lib} is not a known Teem library') from exc
# iteratively find all dependencies and dependencies of dependencies, etc
- oldd = set() # all previously dependencies known
- newd = set([lib]) | set(info['deps']) # newly discovered dependencies
+ oldd = set() # all previously dependencies known
+ newd = set([lib]) | set(info['deps']) # newly discovered dependencies
while oldd != newd:
# while new dependencies were just discovered
tmpd = set()
@@ -202,7 +221,7 @@
tmpd = tmpd | set([lib]) | set(_tlibs[nlb]['deps'])
oldd = newd
newd = tmpd
- tla = tlib_all() # linear array of all libs in dependency order
+ tla = tlib_all() # linear array of all libs in dependency order
# return dependencies sorted in dependency order
ret = sorted(list(newd), key=tla.index)
# exclude "experimental" libraries if not exper
@@ -320,10 +339,11 @@
"""
Given a header file, figures out which lines should be passed to ffi.cdef(), by first
excising comments, and it passing through a very dumb C pre-processor that interprets
- the #ifdef directives to figure out which lines are kept versus ignored.
+ the #ifdef directives to figure out which lines are kept versus ignored, and makes
+ cautious efforts to replace #define'd values in lines that are kept
"""
- def __init__(self, filename: str, defined: list[str], verb: int = 0):
+ def __init__(self, filename: str, defined: dict[str, str], verb: int = 0):
"""
Opens given file and sends it through "unu uncmt", and prepares to parse result
"""
@@ -363,8 +383,8 @@
"""
Do our best to scan input self.ilines to make output self.olines
"""
- for (lnum, linen) in enumerate(self.ilines):
- line = linen.strip() # strip left and right whitespace
+ for lnum, linen in enumerate(self.ilines):
+ line = linen.strip() # strip left and right whitespace
if not line:
continue
if self.verb > 1:
@@ -372,11 +392,11 @@
if line.startswith('#'):
# look for either #ifdef or #ifndef
if match := re.match(r'# *(ifn{0,1}def) +(\S.*)$', line):
- ifdef = 'ifdef' == match.group(1) # else its an ifndef
+ ifdef = 'ifdef' == match.group(1) # else its an ifndef
dname = match.group(2)
if self.verb > 1:
print(f'scan: -------- {match.group(1)} |{dname}|')
- if re.match(r'\S*\s', dname): # if there is whitespace in dname
+ if re.match(r'\S*\s', dname): # if there is whitespace in dname
raise Exception(
f'ScanHdr.scan({self.filename}): line {lnum} |{line}| has '
f'unexpected whitespace in {match.group(1)} argument |{dname}|'
@@ -408,11 +428,12 @@
self.ifstack.pop()
if self.verb > 1:
print(f'scan: -----> ifstack = {self.ifstack}')
- elif match := re.match(r'# *define +(\S+)(\s*\S*)$', line):
+ elif match := re.match(r'# *define +(\S+) +(\S.+)$', line):
+ # elif match := re.match(r"# *define +(\S+)(\s*\S*)$", line):
dname = match.group(1)
dval = match.group(2)
if all(self.ifstack):
- self.defined.append(dname)
+ self.defined[dname] = dval
if self.verb > 1:
print(f'scan: |{dname}|=|{dval}| -------> defined = {self.defined}')
elif re.match(r'# *include', line):
@@ -435,6 +456,23 @@
print(f'scan: -------- DROPPING |{line}|')
continue
# Else we keep the line, and do further processing.
+ # Try to apply #define substitutions
+ for dfn, val in self.defined.items():
+ if not val or not dfn.isupper():
+ # dfn is merely #define'd, but not to anything, or else dfn's letters
+ # are not all upper-case (which is the strong convention for #define's)
+ continue
+ if not '_' in dfn or not len(dfn) > 6:
+ # #define'd name does not contain a '_', or else it is too short, so we
+ # don't consider it more. Seem like an arbitrary limitation, but this is
+ # consistent with the ways that Teem and Teem-adjacent headers #define things.
+ continue
+ # see if dfn appears in line
+ if dfn in line:
+ newline = line.replace(dfn, val)
+ if self.verb > 2:
+ print(f' |{line}| --{dfn}:{val}--> |{newline}|')
+ line = newline
# The parentheses matching of this re is quite fragile
if match := re.match(r'(__attribute__\(.*\))', line):
rpl = line.replace(match.group(1), '')
@@ -494,20 +532,23 @@
f'Missing directory with per-Teem-library biff .csv files {self.path_biffdata}'
)
# This does a lot of error checking
- (self.path_thdr, self.path_tlib, self.have_tlibs, self.exper) = check_path_tinst(
- path_tinst
- )
+ (
+ self.path_thdr,
+ self.path_tlib,
+ self.have_tlibs,
+ self.exper,
+ ) = check_path_tinst(path_tinst)
self.path_tinst = path_tinst
# initialize other members; these will be updated if self.desc() is called to describe
# another library (that depends on Teem) for which we're here to make an extension
# module. The "n" in path_nhdr and path_nlib is for that *N*on-Teem library
- self.path_nhdr = '' # path to header file name.h (nothing to do with .nhdr files!)
- self.path_nlib = '' # path to library file libname.{so,dylib}
- self.libs = ['teem'] # name(s) of libraries the extension module depends on
- self.path_libs = [self.path_tlib] # absolute paths to libraries we depend on
- self.dfnd = [] # things nominally #define'd for sake of cdef()
- self.eca = [] # extra compile args
- self.ela = [] # extra link args
+ self.path_nhdr = '' # path to header file name.h (nothing to do with .nhdr files!)
+ self.path_nlib = '' # path to library file libname.{so,dylib}
+ self.libs = ['teem'] # name(s) of libraries the extension module depends on
+ self.path_libs = [self.path_tlib] # absolute paths to libraries we depend on
+ self.dfnd = {} # things #define'd for sake of header processing for cdef()
+ self.eca = [] # extra compile args
+ self.ela = [] # extra link args
self.source_args = None
self.lib_out = None
if 'teem' == top_tlib:
@@ -528,7 +569,7 @@
self.exper = tlib_experimental(top_tlib)
# create the instance, but don't do anything with it; that depends on other methods
self.ffi = cffi.FFI()
- self.step = 1 # for tracking correct ordering of method calls
+ self.step = 1 # for tracking correct ordering of method calls
def desc(
self,
@@ -535,7 +576,7 @@
name: str,
path_nhdr: str,
path_nlib: str,
- dfnd: list[str],
+ dfnd: dict[str, str],
eca: list[str],
ela: list[str],
):
@@ -545,7 +586,7 @@
:param str name: name of the new non-Teem library
:param str path_nhdr: path to headers for your library, to give to -I when compiling
:param str path_nlib: path to your compile library, to give to -L when compiling
- :param list[str] dfnd: things that should be considered #define'd for ffi.cdef()
+ :param dict[str, str] dfnd: #define's to pass on to ffi.cdef()
:param list[str] eca: for the extra_compile_args parameter to ffi.compile()
:param list[str] ela: for the extra_link_args parameter to ffi.compile()
"""
@@ -567,7 +608,7 @@
raise Exception(f'Need path_nhdr {path_nhdr} to be a directory')
if not os.path.isdir(path_nlib):
raise Exception(f'Need path_nlib {path_nlib} to be a directory')
- self.libs.insert(0, name) # now [name, 'teem']
+ self.libs.insert(0, name) # now [name, 'teem']
self.name = name
self.path_nhdr = os.path.abspath(path_nhdr)
self.path_nlib = os.path.abspath(path_nlib)
@@ -584,7 +625,7 @@
"""
if 1 != self.step:
raise Exception('Expected .cdef() only right after Tffi creation and optional .desc()')
- # want free() available in for freeing biff messages
+ # want free() available for freeing biff messages
self.ffi.cdef('extern void free(void *);')
# read in the relevant Teem cdef/ headers
for lib in tlib_depends(self.top_tlib, self.exper):
@@ -624,7 +665,7 @@
),
# The next arg teaches the extension library about the paths that the dynamic linker
# should look in for other libraries we depend on (the dynamic linker does not know
- # or care about $TEEM_INSTALL). We avoid any reliace on environment variables like
+ # or care about $TEEM_INSTALL). We avoid any reliance on environment variables like
# LD_LIBRARY_PATH on linux or DYLD_LIBRARY_PATH on Mac (on recent Macs the System
# Integrity Protection (SIP) actually disables DYLD_LIBRARY_PATH).
# On linux, paths listed here are passed to -Wl,--enable-new-dtags,-R<dir>
@@ -641,7 +682,7 @@
# https://docs.python.org/3/distutils/apiref.html#distutils.core.Extension
'undef_macros': ['NDEBUG'],
}
- arg1 = f'_{self.name}' # HERE is where we add the leading underscore
+ arg1 = f'_{self.name}' # HERE is where we add the leading underscore
arg2 = f'#include <teem/{self.top_tlib}.h>' + (
'' if self.isteem else f'\n#include <{self.name}.h>'
)
@@ -678,7 +719,7 @@
# We exploit the correspondance between elements of self.libs and self.path_libs
# and HEY that suggests that there should some higher-level object that contains
# all the properties of a library
- for (lidx, lname) in enumerate(self.libs):
+ for lidx, lname in enumerate(self.libs):
cmd = (
f'install_name_tool -change @rpath/lib{lname}.dylib '
f'{self.path_libs[lidx]}/lib{lname}.dylib {self.lib_out}'
@@ -733,13 +774,13 @@
"""
# most common case: return of 1 means there's a biff error
if errval_t.endswith('int') and '1' == errval:
- ret = '_equals_one' # this is defined in lliibb.py
+ ret = '_equals_one' # this is defined in lliibb.py
elif errval_t.endswith('*') and 'NULL' == errval:
- ret = '_equals_null' # this is defined in lliibb.py
+ ret = '_equals_null' # this is defined in lliibb.py
elif 'AIR_NAN' == errval or 'nan' == errval.lower():
ret = '_math.isnan'
else:
- evlist = errval.split('|') # list (likely length-1) of error-indicating return values
+ evlist = errval.split('|') # list (likely length-1) of error-indicating return values
ret = (
'(lambda rv: ' # we make (a string representing) a lambda function
# join, with "or", test expressions for everything in evlist
@@ -758,7 +799,7 @@
"""
if not self.step in (1, 4):
raise Exception('Expected .wrap() only after creation, .desc(), or .compile()')
- biffdatas = [] # a list of rows from .csv files
+ biffdatas = [] # a list of rows from .csv files
for lib in tlib_depends(self.top_tlib, self.exper):
path_bdata = self.path_biffdata + f'/{lib}.csv'
if not os.path.isfile(path_bdata):
@@ -786,7 +827,7 @@
raise Exception("Didn't see wrapper template at {path_lliibb}")
with open(path_lliibb, 'r', encoding='utf-8') as file:
ilines = [line.rstrip() for line in file.readlines()]
- for (lidx, line) in enumerate(ilines):
+ for lidx, line in enumerate(ilines):
# first pass of very simple text transformations we do
# 'lliibb' --> self.name
# 'LLIIBB' --> 'lliibb'
@@ -801,7 +842,7 @@
# bdl[ 0 1 2 3 4 5 ]
# f'{function},{qualtype},{errval},{mubi},{biffkey},{fnln}'
# ---> (rvtf, mubi, bkey, fnln) = _BIFF_DICT[sym_name]
- for bdl in sum(biffdatas, []): # flattening into big list of csv rows
+ for bdl in sum(biffdatas, []): # flattening into big list of csv rows
rvtf = self._rvt_func(bdl[0], bdl[1], bdl[2])
file.write(
f" '{bdl[0]}': ({rvtf}, {bdl[3]}, b'{bdl[4]}', '{bdl[5]}'),\n"
This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.
|
